Data analysis

COVID-19 Analysis

This short report analyzes some of the data from the repository by Johns Hopkins CCSE related to the COVID-19 pandemic.

Preliminaries

We have developed a simple Python module with a set of functions that can be used to visualize the available data. The module is available in the repository (hedera_covid.py), the data as well. We periodically update the datasets with those available online.

# for plotly
from plotly.offline import iplot
from plotly.offline import init_notebook_mode, plot
from IPython.core.display import display, HTML
import plotly as py
import plotly.tools as tls

import numpy as np
from hedera_covid import DataHandler, plot_death_rate, plot_daily_cases, plot_confirmed_cases

# load data
path_confirmed = '../../Data/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv'
path_death = '../../Data/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv'

covid_data = DataHandler(data_confirmed_path = path_confirmed,
                         data_death_path = path_death)

covid_data.confirmed.head()

	Province/State	Country/Region	Lat	Long	...	4/1/20	4/2/20	4/3/20	4/4/20	4/5/20	4/6/20	4/7/20	4/8/20	4/9/20	4/10/20
0	NaN	Afghanistan	33.0000	65.0000	...	237	273	281	299	349	367	423	444	484	521
1	NaN	Albania	41.1533	20.1683	...	259	277	304	333	361	377	383	400	409	416
2	NaN	Algeria	28.0339	1.6596	...	847	986	1171	1251	1320	1423	1468	1572	1666	1761
3	NaN	Andorra	42.5063	1.5218	...	390	428	439	466	501	525	545	564	583	601
4	NaN	Angola	-11.2027	17.8739	...	8	8	8	10	14	16	17	19	19	19

5 rows × 84 columns

First of all, we load the data create an object of type DataHandler. This class can be used to perform internally operation on the dataset, so that data are afterwards ready and clean for plotting.

Initialize list of countries

Now, it is possible to get data of a country to the list that we want to look at. This can be done using

my_country = covid_data.get_country('Name of the country')

The Name of the country is the name given in the third column (Country/Region) of the covid_data dataframe.

For example:

italy = covid_data.get_country('Italy')
italy.keys()

dict_keys(['name', 'dates', 'confirmed', 'deaths', 'daily_new_cases', 'daily_deaths', 'start', 'start_death'])

We can also create a list of countries

covid_data.add_country('Name of the country')

that we can analyze together.

my_countries = ['Italy','Spain','Germany','Austria','France','United Kingdom','US','Sweden','Netherlands']

for c in my_countries:
    covid_data.add_country(c)

Starting from this object, we can now use some functions implemented in our module to visualize the data

Note: (this is an ongoing work!)

Number of confirmed cases

The following figure shows the number of confirmed cases. Note that this is extremely dependent on the different testing protocols and capacity of different countries.

Total

First, we have to collect the data. For the reported cases we use this function:

data = covid_data.get_confirmed_data(start_date=0,n_smooth=7,rescale=True)

Parameter:

start_date to start plotting from a particular day (later than January 22)
n_smooth to smooth the data (7 is usually good)
rescale if you want to rescale curves to the same start (when the number of infected reached 100 in the corresponding country). In this case start_date won't be used.

Then we can create a plotly bar chart (for example) and display it.

init_notebook_mode(connected=True)

fig = {
    "data": data,
    "layout": {"title": {"text": "Confirmed Cases (rescaled in each country)"}}
}

plot(fig, filename = 'figure.html')
display(HTML('figure.html'))

Daily variation

Under the assumption that the number of reported cases is a representative of the total number in each conuntry, looking at the daily new cases can give an idea of whether the countries are flattening the curve.

For this, we can use a function that gather these data for the selected countries.

Parameters:

start_date: day where the plot start (0 = January 22)
n_smooth: smoothing of the data (data will be averaged over n_smooth days, 7 is usually good)
rescale: set True if you want to rescale curves to the same start. In this case start_date won't be used.

We use rescale = True: this means that the curves start when the number of infected (reported) in each country reached 100.

data = covid_data.get_daily_confirmed_data(start_date=0,n_smooth=7,rescale=True)
init_notebook_mode(connected=True)

fig = {
    "data": data,
    "layout": {"title": {"text": "Daily Cases (rescaled)"}}
}

plot(fig, filename = 'figure.html')
display(HTML('figure.html'))

Mortality Rate

We can look at the official mortality rate over time for a selected set of countries

data = covid_data.get_death_rate_data(start_date=30,n_smooth=0,rescale=False)
init_notebook_mode(connected=True)

fig = {
    "data": data,
    "layout": {"title": {"text": "Official mortality rate: # Death/# Confirmed"}}
}

plot(fig, filename = 'figure.html')
display(HTML('figure.html'))

#Parameters:
#* `countries`: a list of countries
#* `start_date`: day where the plot start (0 = January 22)
#* `n_smooth`: smoothing of the data (data will be averaged over `n_smooth` days, 7 is usually good)
#* `rescale`: set `True` if you want to rescale curves to the same *start*. In this case `start_date` won't be used.
#* `log_scale`: set to `True` to scale use logarithmic scale for the *y*-axis
#   
#plot_confirmed_cases(covid_data.countries,start_date=30,n_smooth=0,rescale=False,log_scale=True)